NVIDIA are releasing open source kernel drivers

[rene]

I think the Steam Deck has something to do with this.

Probably. And Intel Arc, in the sense of Nvidia not immediately ceding some cloud-render and/or large-scale CUDA-type applications to Intel.

[Portreve]

Honestly, any development which helps increase Linux's credibility as a alegitimate desktop OS is a good thing.

Separately (and of equal importance) if Nvidia's drivers are open to public scrutiny, we can help fix their deficiencies as well as keep an eye on what the company is (or isn't) doing. As the 1980s Ronald Reagan era expression goes, "Trust, but verify."

I don't trust what I have no capacity to verify.

[sip]

I think the Reds have long been ahead of the Greens in terms of cooperation with Steam ..!
But the Greens are ahead ..!

[Portreve]

So, I was listening to the latest Destination Linux podcast this morning, and as it turns out, the Nvidia driver they're producing for the open source community is actually more oriented towards server use for doing GPU processing. It's presumed that this will eventually lead towards a full driver, but that's not what we're dealing with at the moment.

In fact, as they explained it, this initial driver will not get added to the kernel because it's not written the way the kernel team would want it.

From what Matthew Miller of the Fedora Project was saying, Nvidia would have to take every version of their mainline driver and then rewrite it for Linux, and this would require a team which they would be unlikely to want to fund.

He also as much as said what I've maintained for a long time: if you open-source a driver for something, then the community can get involved and contribute to it just like what happens with the rest of Linux.

I guess we'll just have to wait and see what happens.

[Portreve]

For those who are interested in seeing more about the Nvidia driver, here's the video of the DL podcast I was talking about previously:

Destination Linux #278: Fedora 36 Interview with Matthew Miller

[sip]

Good day.
Today the driver of nvidia 390.151 was updated. As for personal experiences and experiences of the OS and especially downloading videos online and watching them, it has really changed for the better. The video clearly works without fluxes - in the final stage Firefox starts well and quickly ..!
Good job - thank you !!

[rambo919]

https://www.pcgamer.com/nvidias-moved-m ... x-drivers/

[rambo919]

https://linuxstoney.com/status-of-wayla ... a-drivers/

[Hoser Rob]

So, I was listening to the latest Destination Linux podcast this morning, and as it turns out, the Nvidia driver they're producing for the open source community is actually more oriented towards server use for doing GPU processing. It's presumed that this will eventually lead towards a full driver, but that's not what we're dealing with at the moment....

That makes sense actually. Since I'm not a gamer I'm still going to stick to Intel.

[rene]

https://www.pcgamer.com/nvidias-moved-m ... x-drivers/

As to that; please anyone note that except for Richard Stallman by and large no one cares about that, i.e., Nvidia moving proprietary code into firmware; that most certainly I do not. Nvidia's closed kernel-level code in an individual context means that company in one fell swoop controlling what open-source kernel I the individual can run while having my graphics supported -- modulo of course "nouveau" and my freedom to simply not buy Nvidia hardware -- and in a more global context that company in a lot of fell swoops holding back development for example as to Wayland since distributions are and were for some unimaginable reason of the opinion that Nvidia hardware still needed to be supported by them, even if Nvidia made little to no effort the other way around. "Beggars can't be choosers" I guess -- until they simply do,I personally tend to hold, but oh well.

In any case; as long as now the kernel-level interaction details are available through the open-source kernel-level driver then they can as far as I'm concerned hide their precious value-add in optional closed modules, user-space code, firmware or up their butts for all I care about graphics as such; as long as now others can keep this "holding back development" thing from happening I'm golden.

[sip]

https://www.pcgamer.com/nvidias-moved-m ... x-drivers/

As to that; please anyone note that except for Richard Stallman by and large no one cares about that, i.e., Nvidia moving proprietary code into firmware; that most certainly I do not. Nvidia's closed kernel-level code in an individual context means that company in one fell swoop controlling what open-source kernel I the individual can run while having my graphics supported -- modulo of course "nouveau" and my freedom to simply not buy Nvidia hardware -- and in a more global context that company in a lot of fell swoops holding back development for example as to Wayland since distributions are and were for some unimaginable reason of the opinion that Nvidia hardware still needed to be supported by them, even if Nvidia made little to no effort the other way around. "Beggars can't be choosers" I guess -- until they simply do,I personally tend to hold, but oh well.

In any case; as long as now the kernel-level interaction details are available through the open-source kernel-level driver then they can as far as I'm concerned hide their precious value-add in optional closed modules, user-space code, firmware or up their butts for all I care about graphics as such; as long as now others can keep this "holding back development" thing from happening I'm golden.

Good morning. I support you! Everything is correct. The Greens have always led and are not playing fair. !!

[Portreve]

Nah, I'm good. If they aren't willing to release all of it to open source, then it didn't really interest me much. Besides, there's alternatives out there, so this is Nvidia's game to lose.

[sip]

Good day. For me personally, as a Linux Mint user and a Linux user in general, this is very important. Linux Mint is my main system both at home and in business. Until 2009, I worked on two Windows and Linux systems. After 2011, I switched to Mint and Debian. It is important to me that the system is open source, secure and stable. I think it's not only important to me. I am sure that people who know and are responsible for such cooperation on the part of Linux kernel developers will sit down at the negotiating table to resolve this issue.

[rene]

The firmware <-> software boundary is in this context really not the one to worry about; the "hard- or firmware" <-> software one is. That is...

The dividing line between hardware and firmware is here fluid in the first place. I mean, the "hardware", the chips as such, are in the conceptual sense no other than programmatic logic implemented in physical logic gates directly rather than via some kind of in physical logic gates implemented microprocessor/controller that "maps" that same programmatic logic, the firmware, onto lower-level physical logic gates.

Both hardware and firmware in this sense are simply "deviceware": programmatic logic that is and stays on the device as such (and having e.g. through software user-upgradeable firmware for a device does not change that in the slightest: potentially helps said user by making its "hardware", its device, upgradeable rather than necessarily always-buggy-if-initially-buggy and/or upgradable to newer functions/standards).

It is the line between this deviceware on the one and software on the other where the issue of Nvidia's closed kernel-module in the context of the Linux kernel resides. I.e., at code that runs not on the device itself but on your general purpose, central microprocessor; as such necessarily interferes with and/or influences any other code running on said CPU. The x86-64 CPUs that we are using have only two modes available; kernel-mode in which code can do any and all and user-mode in which code can do little to nothing (as to hardware programming at least) without kernel-mode code needing to give the go-ahead anyway; 32-bit x86 had two additional "rings" in between but these due foremost to something precisely like a GPU-driver needing full-access anyway for performance reasons were done away with for 64-bit.

This is to say that the Nvidia closed, binary kernel driver can do literally any and all; interfere with any and all and for the paranoically inclined in theory even act maliciously and/or harness capabilities of all the rest of your computer. It's to say that at the very least conceptually the millions upon millions of lines of open, reviewable Linux-kernel code in the one fell swoop of loading the "nvidia" kernel module turns itself into a non-reviewable black box of Microsoft-esque proportions.

In this case "at the very least conceptually" is to moreover say exactly that and not "not practically". Practically you on kernel-level need to be able to look at something that's as big a part of kernel-level code-flow as a GPU driver and this is then also the reason why e.g. at some point the "tainted" flag in kernel error messages was implemented; to show kernel developers that at the time of a reported bug closed (and/or later otherwise problematic) code was loaded; code they can no longer review for purposes of finding the problem. To basically say, "Take your trouble to Nvidia because we no longer know what you are running."

TL;DR:

As long as Nvidia's closed code stays on Nvidia's device Nvidia may feel perfectly free to <bleep> over said Nvidia device. It's where Nvidia's closed code interferes with anything else -- or in this case even everything else -- where the line must be drawn.

Yes, due to logic-wise fluid nature of hardware/firmware/software boundaries it's not impossible to have open software still cause issue due to closed hardware/firmware and I'd not be surprised if some surfaced even in practice while the Nvidia kernel-level code is beaten into shape for integration in the Linux kernel and/or while "nouveau" is being improved due to it. And, also yes, from the perspective of full-disclosure-or-die and/or nerd-level interest everything open is better -- but I personally still don't give a shit largely about graphics as such, certainly not about Nvidia in particular, and am perfectly fine as long as now having an open kernel-level driver does not interfere any more with other development.

I always see tons of typo/grammos just after posting but I'm not going to review this essay, so there; live with it: Nvidia has even tainted my bugs now...

[Hoser Rob]

https://www.pcgamer.com/nvidias-moved-m ... x-drivers/

As to that; please anyone note that except for Richard Stallman by and large no one cares about that, i.e., Nvidia moving proprietary code into firmware...

IMHO RJS is a blithering idiot overgrown adolescent. Intel has been doing this with their open source drivers for years. I wonder how many RJS loving FOSS nut Linux users know that their machines would not run Linux without that firmware ? Well, if they installed Debian they'd find out pretty quickly.

[rene]

He goes by "rms" rather than "RJS" and I'll note that it's to be sort of at least conceptually appreciated him having managed to turn Asperger's syndrome into an international movement. For example Hans Reiser only managed to turn same into life in prison...

[Portreve]

I'm not sure what constructive purpose is served by calling Richard Stallman a blithering idiot. While it is true that there's times when it's necessary to cohabitate with proprietary closed source type operations and entities, the public interest is not served (or not particularly well served) by having and/or encouraging the closed-source development model.

Moreover, what have you been doing to better mankind at large and promote the general welfare?

If there's anything that personal experience, observing the behaviors of both companies and government institutions, and whistleblowing, has shown us, it's that there's no valid reason to blindly trust others, which is exactly what one does every single time one uses closed-source anything. We get lucky every time we're using something proprietary and it doesn't actually do something bad to us.

[rene]

I'm not sure what constructive purpose is served by calling Richard Stallman a blithering idiot.

Heck, many people are not even sure what constructive purpose is served by Richard Stallman, period.

Any case; see above firmware vs software thingy as to how at least this sub-thread, the one as to Nvidia moving closed logic into firmware before opening up, is not about open/closed code as such but about open or closed code being able to hurt anything other than itself. It's fairly well-established human doctrine to have freedom to do as one wants be considered inversely proportional to potential for damage to others; in this case to have Nvidia's freedom to keep its so conceived value-add secret through firmware be considered its prerogative. You (and I) can after all still decide to simply not buy their hardware if we don't like it.

[DisturbedDragon]

Good news to be sure. Never understood most Linux users' animosity toward Nvidia. I've used Nvidia cards exclusively from the old mx440 to my current RTX2080 Super without issue.

[Hoser Rob]

Good news to be sure. Never understood most Linux users' animosity toward Nvidia. I've used Nvidia cards exclusively from the old mx440 to my current RTX2080 Super without issue.

Well, just take a look elsewhere and note which brand of GPU generates the most issues posted here.